Today in Analytics...

Lotta Juice

Ramblin' Wreck
Messages
511

Thought this was an interesting visual showing differences between yards per play and expected points added rankings.
GT is above average in yards per play and just below average for EPA. Seems to reflect what we've seen on the field with an ability to move the ball at times but some struggles to turn it into points.

Next 2 games will be wildly different offenses as you can see the opposite ends of the graph uva and vt are situated. Pitt is way up in that top right corner. UNC is up there too. Duke actually doesn't look too bad per this graph. Notre Dame and VT are bottom left. BC is just sitting in the middle

👀👀Clemson.
 

ibeattetris

Helluva Engineer
Messages
2,584

Thought this was an interesting visual showing differences between yards per play and expected points added rankings.
GT is above average in yards per play and just below average for EPA. Seems to reflect what we've seen on the field with an ability to move the ball at times but some struggles to turn it into points.

Next 2 games will be wildly different offenses as you can see the opposite ends of the graph uva and vt are situated. Pitt is way up in that top right corner. UNC is up there too. Duke actually doesn't look too bad per this graph. Notre Dame and VT are bottom left. BC is just sitting in the middle

👀👀Clemson.
Yep. Driving the length of the field will be positive yards. Getting to the 1 and kicking a field goal is negative EPA as you're expected to score more than 3. As the tweet says, it puts context around the value of every play run. Cool tweet.
 

GTpdm

Helluva Engineer
Messages
1,451
Location
Atlanta GA

Thought this was an interesting visual showing differences between yards per play and expected points added rankings.
GT is above average in yards per play and just below average for EPA. Seems to reflect what we've seen on the field with an ability to move the ball at times but some struggles to turn it into points.

Next 2 games will be wildly different offenses as you can see the opposite ends of the graph uva and vt are situated. Pitt is way up in that top right corner. UNC is up there too. Duke actually doesn't look too bad per this graph. Notre Dame and VT are bottom left. BC is just sitting in the middle

👀👀Clemson.
Interesting that Army is one of the off-diagonal teams on the "good" (?) side of the graph: middle-of the road YPP rank but pretty darn good EPA rank.
 

GTpdm

Helluva Engineer
Messages
1,451
Location
Atlanta GA
Maybe their defense has gotten them a short field often enough to sway the numbers.
Interesting point. How much is EPA influenced by having a short field? If it plays a role at all, then EPA doesn't really measure offense 100%, because the defense and S/T giving you good field position would tend to inflate your EPA, all other things being equal.

Can someone more familiar with how EPA is calculated shed some light, here?
 

Lotta Juice

Ramblin' Wreck
Messages
511
Interesting point. How much is EPA influenced by having a short field? If it plays a role at all, then EPA doesn't really measure offense 100%, because the defense and S/T giving you good field position would tend to inflate your EPA, all other things being equal.

Can someone more familiar with how EPA is calculated shed some light, here?
I don't think EPA is influenced by having a short field. It certainly accounts for field position in the calculations but it's more the delta between the EPA before a play happens and then the result so it's more tied to what was achieved on the play.

I think Army's success in EPA and mediocrity on YPP is probably that they just don't have a lot of explosive plays which could limit their yards per play but they are successful in moving the ball and not hurting themselves so they are above average on EPA. I'm not going to claim to be an expert but that's my take.

If you look at the stats below it shows Army ranked at 126 out of 129 in "Long Scrimmage Plays".
http://www.cfbstats.com/2021/leader/national/team/offense/split01/category30/sort01.html

And take a look at who is no. 1.....our next opponent, UVA.
 

ibeattetris

Helluva Engineer
Messages
2,584
Interesting point. How much is EPA influenced by having a short field? If it plays a role at all, then EPA doesn't really measure offense 100%, because the defense and S/T giving you good field position would tend to inflate your EPA, all other things being equal.

Can someone more familiar with how EPA is calculated shed some light, here?
Here is a discussion on it: https://saiemgilani.github.io/cfbfa...xpected-points-model-fundamentals-part-i.html

This graph shows how epa is affected by down and distance
1634247787086.png


EPA would be influenced by field position. Take for example a contrived example where your starting field position is always on the 1 yard line of your opponents side and you score a 1 yard TD every play. The difference between that one td and a one yard gain at another field position is massive.

I also could not tell from the graph, but epa is both an offense and defensive stat. Your defense forcing a turnover is a +epa event (as is a sack etc.).
 

ibeattetris

Helluva Engineer
Messages
2,584
Just because I couldn't resist, here is 2014:
1634250410881.png

Offenses are in general pretty high on the epa/play this year. I used the same filter of epa/play > 0.038 for each graph. There are 34 more teams this year averaging over that amount, and 11 more teams averaging over 0.2. We were solidly in the top 3, but would only be in a tight pack with top 10 this year.
 
Top